Content aware fill based on similar images

A digital image processing method comprising:

• receiving a first input that identifies a source image from amongst a plurality of images stored in an image repository;
• receiving a second input that defines a selection mask within the source image, the selection mask encompassing an object for removal that appears in the source image;
• identifying a plurality of images that are related to the source image;
• selecting one of the plurality of related images as a best-match related image;
• removing a portion of the source image that is defined by the selection mask;
• and
• filling in the removed portion of the source image using a segment of the best-match related image.

A non-transitory computer readable medium encoded with instructions that, when executed by one or more processors, causes a digital image manipulation process to be carried out, the process comprising:

• receiving a source image;
• in response to user input, defining a selection mask within the source image, the selection mask encompassing an object for removal that appears in the source image;
• defining m matching masks within the source image, each of which comprises an area positioned adjacent to the selection mask;
• identifying n related images that are to be compared to the source image;
• counting a number of features R(n′) that appear in both the selection mask and an n'th one of the related images;
• counting a number of features E(n′, m′) that appear in both an m'th one of the matching masks and the n′ related image;
• selecting one of the n related images as a best-match related image based on a matching score that depends on E(n′, m′) and R(n′);
• and
• replacing a portion of the source image that is included in the selection mask with an image segment derived from the best-match related image.